Digital Scoring of Stromal Tumour Infiltrating Lymphocytes in Triple Negative Breast Cancer

Abstract

Accurate assessment of tumor-infiltrating lymphocytes (TILs), particularly in the stromal region, is a valuable prognostic and predictive biomarker in triple-negative breast cancer (TNBC). However, traditional scoring methods are subjective, labor-intensive, and lack standardization. This paper presents a computational pipeline for automated and quantitative evaluation of stromal TILs from whole slide images (WSIs) of pre surgery needle core biopsies of TNBC patients. The pipeline employs a multi-step approach, combining traditional image analysis techniques with deep learning models. By combining traditional image analysis and deep learning approaches, the pipeline leverages the strengths of both methodologies to overcome the challenges of TIL quantification, including tissue heterogeneity and variability in TIL appearance and distribution. This work represents a significant step towards reliable and scalable TIL quantification, facilitating large-scale studies and enabling personalized treatment strategies based on TIL biomarkers.